Based on original by Ara Howard @drawohara: https://github.com/dojo4/dojo4/blob/master/docs/how-tos/how-to-estimate-a-project.md

How to estimate a project

Do

• Estimate broad categories of work such as ‘api’, and ‘site design’.
• Avoid estimating features and focus on estimating business goals and requirements such as ‘can buy a product’ vs ‘integrate with stripe’, or ‘ability to download movies’ vs ‘ability to click a red button and download a movie if no other movie is in the queue which is displayed in a widget in the upper right corner.’
• Keep estimates to one page, or even one paragraph with bullet points. If it doesn’t fit on a page it’s likely the project will cost > 100k and you’ll want help, a white board, and 1/2 day to make the estimate. Never do these kind for free.
• Ask for an internal review of your estimate by one team member.
• Work in teams to estimate large projects.
• Consider collateral work. For instance: a logo implies a favicon and an apple-home-screen icon, a production deploy implies a staging deploy, monitoring, and database backups… think from the client perspective about everything they will need to do to be live including DNS procurement, password management, account setup, etc.
• Consider the personality of the client. How much communication do they need? How much detail in tickets/invoices/etc? If they will be involved multiply by about 200%, if they want to help 300%.
• Time box your effort estimating. 15 minutes is good. 1 hour is pushing it. Too many unknowns to do this suggests a paid exploration period of 1-3 days in order to write code, research, and provide a more accurate estimate. Alternatively give big ranges - ask for paid exploration to narrow them.
• Look for something small to start. If 15k of work is clear and the remaining 85k is murky, suggest getting going tomorrow on the clear discrete bit and returning to estimate the murky bit when more it known. The client will still want a ballpark - but this can unblock negotiations.
• Avoid talking about rates in estimates. Ideally we can charge 25k for completed project regardless of how many hours were spent when that number was in the range estimated (this doesn’t happen often but is possible).
• Include a minimum and maximum number of hours we’ll work on the project regardless of its completion. This protects us from never-ending feature creep and the client by promising a minimum effort level. To calculate this divide the highest cost you can imagine by ~$100 (we are losing money) for the maximum and, for the minimum, divide the lowest cost you can image by ~$200 (we are killing it).
• Include a suggested start date and end date. Assuming the team is working at 50% capacity to get this number. So if you think it’ll take 2 people working 30 hrs/wk for one month suggest an end date 2 months after the start date.
• Include an expiration date for the estimate. When a client comes back 3 months later ready to start something any research done will need to be repeated, now apis may exist, the product may have many competitors where once it had none, our pipeline may be more full or empty than it was when we originally made the estimate… all of these things can cause a big change to an estimate.
• Consider our pipeline. When it is empty, aim for smaller bits of cheaper work. When it is full aim high.
• Suggest staying on our current stacks. Moving to another framework or deployment container always need considered.
• State the problem you’ll be solving in the estimate. Are you sure they understand what they are asking for?

Do Not

• Underestimate the importances of having access to all production content, apis, etc up front. Not having these up front has the ability to make any estimate a wild ass guess from the get go.
• Underestimate the lag time legal can add to a possible start date. This goes along with getting any contract signed ASAP.
• Estimate specific details unless every single aspect of it is known (it never is).
• Underestimate authentication. If the project colors outside the lines (we don’t already have it 100% programmed), it will cost at least 2.5k to get right if it is a standard 3rd party (OAuth, Google, etc.). If it’s custom it will absolutely consume a week before the project is finished.
• Underestimate deployment and delivery. A day is 1k. Even getting a DNS entry setup and static files to a client takes time and energy.
• Underestimate integrations. An external api or dependency adds an order of magnitude of difficulty because of communication, bugs, blocking, and changes.
• Underestimate the cost of duration and handoffs. A project that stretches over long periods of time and has many handoffs incurs lots of increased costs. For small projects (< 10k) expect this to be a factor of about 100%.

An Example

Hi Clientperson,

I’ve reviewed your fucking vague Microsoft Word doc full of mutually contradictory requirements docs and see the following:

You will need:

• A micro site put up on http://sweet-new-thang.com.
• An API integration with http://awesome-sauce.com in order to conquer the world and sell lots of kittens.
• Some sort of ability to turn this into books full of kittens.

We see the following next steps:

• Getting your micro-site up is something we can start next Monday. This could take 3-5 man days and cost ~3.5-4k.
• The api integration with http://awesome-sauce.com will take a bit of research since it’s currently in beta and the repo seems to be under active development. We’d like to do a time-boxed assessment of it now in order to fully understand the difficulty in integrating with it in production, although we think this will take about 2 weeks, or < ~10k.
• The book fulfillment is a really big project and we’re not prepared to estimate it at this time without access to the fulfillment api that is being built by http://aint-going-to-happen-any-time-soon.com and without reviewing at least three samples of complete book content – the animated gif of the kitten isn’t cutting it… In any case, while it is possible that this could cost 19 bucks, as you suggest, it also may cost $999,999 bucks and we think it’s prudent to consider this effort in another sprint.

Assuming we can start work next Monday, and no later than two weeks from now, we’d be able to finish the microsite in 2 weeks and have a go/no-go on the awesome-sauce API in that same time frame. assuming our research on the API doesn’t reveal any blockers, we’d finish the integration 6 weeks from now.

Ignoring the book fulfillment you’ll be looking at:

• ~ 5-14k
• ~ 6 calendar weeks until launch

If this sounds good let us know asap and we’ll get this up in concordnow.com where we can continue to make adjustments.

Cheers - I’m sure you’ll rule the world with kitten books.